Traffic Signal Control with Connected and Autonomous Vehicles

The accelerating pace of advances in sensing, communications, and computation provides significant opportunities for enhancing mobility and safety in the transportation system. Many challenges, however, must be tackled for a smooth transition in the deployment of such technologies. This research develops an Intelligent Intersection Control System (IICS) for both Connected/Automated Vehicles (CAVs) and conventional vehicles in the traffic stream.

The IICS framework operates an isolated intersection within the detection range from the stop bar based on input arrival data. These arrival data are obtained through radars (which is used to obtain information regarding the arrival of both AVs and conventional vehicles) and Dedicated Short-Range Communication (DSRC - which is used to both receive arrival information and to send optimized trajectories to CAVs). At the core of IICS, an optimization algorithm minimizes the delay at the intersection by making real-time decisions on Signal Phase and Timing (SPaT) alongside CAV movements. Then an interface code to the signal receives the optimized SPaT decision to be transmitted to the signal controller in the cabinet, which operates the signal displays according to the optimized plan. In parallel, the CAVs receive readable instructions regarding their optimal trajectories.

The optimization algorithm was initially implemented in MATLAB. The simulation results were compared to the performance of the same intersection with conventional traffic, and they show, that depending on the saturation headway at the stop bar, the algorithm results in 38 to 52 percent reduction in average travel time. We implemented the IICS at FDOT's TERL laboratory facilities and tested it under different scenarios over the course of a two-day experiment with six vehicles. The outputs and video footage from the test showed this is a promising direction, as the system is capable of providing optimal trajectories to CVs and AVs in order to reduce delay at isolated intersections due to unnecessary stops (visual results at http://avian.essie.ufl.edu/).

We are currently extending the optimization algorithm to be able to handle high traffic volumes and oversaturated conditions, which require keeping track of each vehicle over multiple cycles. The code is also being transferred to Python, to accelerate its performance, and prepare it for field testing.